Pricing

Same agent. Same governance. Same provenance. Four price points.

Krentix isn’t sold by the token or by the seat. Pick a power tier — the architecture stays identical (twelve verifiers across seven providers, constitutional governance, full provenance trail), only the model lineup underneath changes. Switch tiers per session, or per request. The dial is the product.

Zero

$0/ day forever

Hobbyist · Student · Free trial

  • All 12 personas active on free providers
  • Cerebras: Qwen 3 235B, GPT‑OSS‑120B, GLM‑4.7
  • OpenRouter, Moonshot Kimi, Together, Mistral free tiers
  • Ollama if installed locally (no rate limits)
  • Real twelve-verifier ensemble, just on free weights
  • Source Mode included
Open the agent →

Low

$0.05/ day

Freelancer · Side projects

  • Free Cerebras tier on most personas
  • Claude Haiku 4.6 on the verifier roles
  • Gemini 2.5 Flash on architecture · UX
  • ~10× cheaper than always running max-effort
  • Same verification quorum as higher tiers
  • Source Mode included
Try it →

Medium default

$0.20–0.60/day

Standard professional

  • Claude Opus 4.7 on Code Skeptic
  • Sonnet 4.6 on Constitutional Tribune (governance)
  • GPT‑4o on security · UX
  • DeepSeek on regression · risk
  • Free Cerebras OSS for diversity (3 personas)
  • Frontier-class outputs at ~1/10 always-max-effort cost
Open →

Full power

$5–15/day

Production critical · Leaderboard runs

  • Seven distinct labs voting per answer
  • Claude Opus 4.7 (4 personas) · Grok 4 (xAI)
  • DeepSeek V4 Pro + Qwen 3.5 397B (via Together)
  • Gemini 2.5 Pro · GPT‑4o · Mistral Large
  • Kimi K2 Thinking (Moonshot)
  • Extended thinking on governance personas
Run leaderboard mode →

The dial vs. everyone else.

Cursor, Copilot, Devin, every other major coding agent: pick one model, pay maximum on every keystroke, no throttle. Krentix is the only one that gives the user the curve.

Capability Krentix Cursor Copilot Devin
Switch model lineup per session Yes · 4 tiers, 1 API call Pick base model No No
$0/day forever lane Yes · 5 free providers No (Pro $20/mo) No (Free tier limited) No ($500/mo)
Multiple labs verify each answer Yes · 7 labs voting One model One model Single agent loop
Source-bounded answers (NotebookLM-style) Yes · with citations RAG over codebase Workspace context No
Per-answer provenance trail Yes · which model voted what No No Limited
Public benchmark harness in repo Yes · HumanEval · MBPP+ No No Self-published
Constitutional governance / veto Yes · Tribune + E-R protocol No No No